A new Ai txt2img model for creating pics came out in August 2024.

Flux, it is NOT censored, and it can be run locally. It performs much better than SD1.5 and SDXL. It's on par with Midjourney and Dall-E 3, but it is local, so no prompt is censored.

You can try it for free here: https://huggingface.co/spaces/black-forest-labs/FLUX.1-dev

It can be run on as little as 4GB of VRAM (it is tricky to do but possible), so if you have run SDXL or SD1.5 or pony you can run this one.


I have made some LoRAs for this model. A hanging model and a model where a girl is sucking the barrel of a gun.

I'm still learning how ot properly make LoRAs so those are versions 0.5 I will make a better version in the future.


1. Hanging LoRA, trained mostly on 3d images, so it makes everything look 3D, weight 0.5-1, trigger words hanging, hang, noose


https://files.catbox.moe/oa69r0.safetensors


2. Gun in the mouth LoRA, sometimes the gun is not drawn properly.


https://files.catbox.moe/qyyppb.safetensors



Both have around a 50% success rate. They combine well with the character LoRAs.

Have fun, and if you use them, please post results.

If anyone has a specific fetish, please provide a dataset and i will train it.


It doesn't do anime well yet. SO I will post only the 3D pic here.